Most Trusted. Most Cited. Most Read

PiscoMed Publishing Welcome cooperation


Spectral Unmixing of Hyperspectral Images in the Presence of Small Targets

Over the last decades, hyperspectral imaging, also known as imaging spectroscopy, has faced a growing interest in multiple fields such as astronomy, agronomy, military, geography. In hyperspectral images each pixel is acquired by recording images across numerous spectral bands from visible light to infrared, such as the acquired information is an approximation of the reflection spectra of the imaged scene. Given the spatial resolution of the current hyperspectral sensor, a pixel generally contains multiple materials. As each material has a given reflection spectrum, the observed spectrum is a mixture of those materialsspectra called endmembers.


Sylvain Ravel, Caroline Fossati and Salah Bourennane conducted an assessment of the spectral unmixing of hyperspectral images in the presence of small targets. The survey result published on the of Remote Sensing.


In this paper, they consider a linear mixing model where the pixels are linear combinations of those reflectance spectra, called endmembers, and linear coefficients corresponding to their abundances. An important issue in hyperspectral imagery consists in unmixing those pixels to retrieve the endmembers and their corresponding abundances. They consider the unmixing issue in the presence of small targets, that is, their endmembers are only contained in few pixels of the image. They introduce a thresholding method relying on Non-negative Matrix Factorization to detect pixels containing rare endmembers. They propose two resampling methods based on bootstrap for spectral unmixing of hyperspectral images to retrieve both the dominant and rare endmembers.


Their experimental results on both simulated and real world data demonstrate the efficiency of the proposed method to estimate correctly all the endmembers present in hyperspectral images, in particular the rare endmembers.


For more information, please visit: